1. Helsinki 19 May 2006 Fine Protection of Data-Paths in Multi-Layer Networks Based on the GMPLS paradigm G.Oriolo, Università Tor Vergata, Roma joint work with P.Iovanna, Coritel & Ericsson Lab, Roma R.Sabella, Coritel & Ericsson Lab, Roma L.Sanità, Università Tor Vergata, Roma CoRiTeL Università Tor Vergata

2. Outline CoRiTeL New Generation Optical Networks Protection Strategy Optimization Problem Heuristic Combinatorial Algorithm Simulations and Results

3. New Generation Optical Networks Key Issues:  Rapid growth of Internet traffic  Several types of traffic with different QoS Objectives:  Flexibility and Efficiency  Handling different QoS and different levels of protection against failures CoRiTeL

4. Multi-Layer Network Scenario MPLS Logical Layer WDM Physical Layer LSR OXC Traffic Requests (lsp-s) Optical fiber GMPLS Control Plane CoRiTeL Ligthpath

5. The Protection Problem  Given a multilayer network and a set of traffic requests Set up a set of lightpaths and assign to each HP traffic request a logical path, i.e. a sequence of lightpaths, for the working scenario… Working Path Source Destination Back-up Path CoRiTeL  … and a logical path for each failure scenario (back-up paths)

6. End to End Protection  Assign to each traffic request one working path and one back-up path that are disjoint CoRiTeL  Wei Wei, Qingji Zeng, Yun Wang, “Multi-layer Differentiated Integrated Survivability for Optical Internet” Photonic Network Communications, vol 8, number 3, pp 267-284, 2004  Qin Zheng, G.Mohan “Protection Approaches for Dynamic Traffic in IP/PMLS-over-WDM Network” Communications Magazine, IEEE, vol 41, pp 24-29, 2003  Yinghua Ye, Assi C., Dixit S., Ali M.A, “A Simple Dynamic Integrated provisioning/protection scheme in IP over WDM networks”, IEEE Communications Magazine, Nov 2001 – pp 174-182

7.  We associate a scenario with each failure event + one for the normal state  We want to assign to each traffic request one working path and one back-up path for each possible failure scenario Protection Strategy  Failure event: single optical link failure (causing multiple logical failures)  The protection strategy can be directly extended as to deal with different types of failures (as node ports) CoRiTeL  … therefore we want to assign to each traffic request a (logical) path for each possible scenario

8. Protection Strategy 2 Recovery Base Lightpath: we reconnect the source and the destination of the failed lightpath while keeping fixed the rest of the logical path CoRiTeL Advantages:  confines signaling to the portion of the network that has failed  saves time of processing and reaction  makes easier fault management

9. Protection Strategy 3 A traffic request is diverted from its working path p only if some lightpath of p fails CoRiTeL  Advantage: in each scenario resource sharing is maximixed A traffic request might be diverted to different back-up paths when two edges e,f of a same lightpath of p fails

10. Protection Strategy 4 Two traffic requests sharing a same lightpath j on their working paths may have different back-up paths when some edge e of j fails CoRiTeL LSP 1 LSP 2  Advantage: in each scenario resource sharing is maximixed

11. More Assumptions After a failure is repaired, working paths are reinstated (revertive strategy) Lightpaths must be set-up in advance; still a lightpath may carry different traffic requests in different scenarios

12. Feasibility of a solution (logical layer) Consistency check Scenario 0Scenario 1Scenario m LSP 1 LSP 2 LSP k l 1 -l 2 -l 3 l 1 -l 5 -l 3 l 4 -l 6 l4-l1-l7l4-l1-l7 l 1 -l 2........................ Capacity check CoRiTeL 0 123 Bandwidth on lightpath l 1 Scenario LSP 2 LSP 3 LSP 4 4 Failure!!! LSP 1 Wavelenght capacity

13. Feasibility of a solution (optical layer) CoRiTeL Consistency check Capacity check Physical link 1 Lightpath 1 Lightpath 2 Lightpath p........................ Physical link 2 Physical link m yes no yes no yes Intra-ports LSR 1 Intra-ports LSR 2 Intra-ports LSR n yes no

14. The Off-line Protection Problem Given: Find: As to:  A Multilayer Network  A set of HP traffic requests (LSP-s) k = 1 … K, with source s(k) destination d(k) and bandwidth b(k)  The Logical Topology  Working Paths for each traffic request  Back-up Paths for each traffic request, for each scenario  Serve as many traffic requests as possible CoRiTeL

15. An ILP formulation 1  N el := the set of LSR-s nodes  N opt := the set of OXC-nodes  E opt := the set of (directed) optical links  E port := the set of (directed) intra-ports links  n opt (i,j): = the number of wavelengths in (i,j) E opt  n port (u,v):= the number of incoming/outgoing intra-ports in (u,v) E port  w = bandwidth available on a wavelength/intra-port  K := the set of LSP-s  s k - t k - b k := respectively LSR-source, LSR-destination and requested bandwidth of k K  S := {s 0, …, s m } the set of failure scenarios. s 0 is the working scenario, m = |E opt |/2  L := the set of candidate lightpaths to be set up  L opt (i,j):= the set of lightpaths L using (i,j) E opt  L port (u,v):= the set of lightpaths L using (u,v) E port  L + (u) := the set of lightpaths L starting from node u N el  L - (u) := the set of lightpaths L ending in node u N el L(s) := the set of lightpaths L failing in the scenario s S (with L(s 0 ) = Ø) Define: CoRiTeL

16. An ILP formulation 2 Decision Variables: CoRiTeL

17. An ILP formulation 3 The objective is to maximize the number of served LSP-s (1) The total number of lightpaths using the optical link (i,j) does not exceed the number of available wavelengths (2) The total number of lightpaths using an intra-port between LSR u and OXC v does not exceed the number of available intra-ports (3) Flow-constraints for LSP k in scenario s (4) The total amount of traffic passing through each lightpath in each scenario does not exceed the capacity of the lightpath itself (5) If k is routed on a lightpath l in the working scenario, it must be routed on l in each scenario l is active (6) k cannot be routed on a lightpath l that is not active in a scenario s (7) For sake of shortness we assume that a set L of candidate lightpaths is given CoRiTeL

18. Heuristic Approach Sequential procedure based on successive shortest paths computations: - simple and easy to implement - can be extended to on-line HP traffic Provisioning Step Recovery Step Local Search Step Working paths research Local search to improve the solution Back-up paths research Lightpaths are set-up dynamically while searching for the routing of an LSP CoRiTeL

19. Provisioning Step  LSP-s are sorted according to non decreasing values of the minimum bandwidth they need, i.e. MinHop(k)*b(k)  For each LSP k the Routing Engine searches a working path, taking into account the working paths and the set of lightpaths set-up so far  If such a path is not found, then LSP k is temporary discarded CoRiTeL

20. Recovery Step  Failure scenarios are sorted according to the loads on each optical link induced by the working paths  For each failure scenario, LSP-s affected by the failure are sorted according to non decreasing values of the minimum bandwidth they need  For each LSP k the Routing Engine searches an alternative path for the failed lightpath, taking into account: the working paths, the back-up paths and the set of lightpaths estblished so far  If such a path is not found, then the LSP is temporary discarded CoRiTeL

21. Local Search Step  We try to accommodate LSP-s that have been discarded so far  This is done by simple swapping rules and by means of the Routing Engine CoRiTeL

22. Routing Engine It uses Dijkstra’s algorithm on an auxiliary network composed by:  Two types of nodes 1. OXC-s 2. LSR-s  Three types of links 1. optical links 2. intra-ports 3. lightpaths set up so far and it may establish new lightpaths CoRiTeL

23. 1)W lightpath = # optical hops lightpaths links Link weights 2) optical links if λ a ij > 1 if λ a ij = 1 3) intra-ports links if λ a ij >1 if λ a ij =1 CoRiTeL

24. Alternative Strategies for Simulation Objective : comparing our strategy wrt to other possible strategies Up to now: comparing our strategy wrt to 2 less refined strategies defined by: 1) forcing an LSP to use the same backup path for the failure of edges in a same lightpath CoRiTeL

25. Objective: comparing our strategy wrt to other possible strategies Up to now: comparing our strategy wrt to 2 less refined strategies defined by: 2) forcing two LSP-s sharing a lightpath j to share also the back-up path when an edge e of j fails CoRiTeL LSP 1 LSP 2 Alternative Strategies for simulation

26. Simulation Network: 19 nodes 36 links Hypothesis  Each OXC connected with an LSR  Each optical link is bi- directional  Number of intra-ports = 70% number of λ crossing the OXC  w λ =2,5 Gb/s CoRiTeL k C tot b(k)b(k) min_hop(k)   Network load

27. Results Network Load ρ 10%100%0 20%100%0 30%96%0 40%87%44% 50%60%51% 60%41%39% 70%30%29% 80%25%26%  π = percentage of lost traffic by the main strategy Percentage Lost Traffic vs ρ 0% 10% 20% 30% 40% 50% 0102030405060708090 Network load ρ Lost Traffic π π1π1 π2π2  π 1 = percentage of lost traffic by the alternative strategy (1)  π 2 = percentage of lost traffic by the alternative strategy (2) CoRiTeL

28. Conclusions  Fine Routing / Protection strategy for MPLS over WDM network via a simple shortest path algorithm  Main elements of the protection strategy: scenario dependency, LSP granularity, ligthpath recovery base, high resource sharing  First computational tests show the advantages of our strategy Future Work  Exact comparison with End to End protection strategies via ILP techniques CoRiTeL